anytype-mcp

Automatically transforms Anytype's OpenAPI specification into MCP tool definitions at runtime using the OpenAPIToMCPConverter component. This eliminates manual tool definition maintenance by dynamically generating tool schemas, descriptions, and parameter mappings from the source OpenAPI spec, ensuring AI assistants always have access to the latest API endpoints without code changes.

The MCPProxy component implements the MCP protocol specification, handling incoming tool listing requests and tool execution calls from AI assistants. It translates MCP-formatted requests into HTTP calls to the Anytype API via the HttpClient layer, manages response serialization back to MCP format, and handles protocol-level error mapping to ensure AI assistants receive properly formatted results.

Supports efficient bulk operations on multiple objects through MCP, allowing AI assistants to update properties, apply tags, or modify relationships across many objects in a single workflow. Rather than making individual API calls per object, batch operations reduce latency and improve efficiency when AI needs to perform coordinated changes across the knowledge base.

Anytype's architecture ensures all data is encrypted locally before any network transmission, and the MCP server respects this encryption model. Objects are stored encrypted in Anytype's local database, and when accessed through the API, decryption happens locally before data is returned. The MCP server does not handle encryption/decryption directly — it relies on Anytype's local client to manage keys and encryption, ensuring end-to-end encryption even when accessed through AI assistants.

The HttpClient component manages all HTTP communication with the Anytype REST API, handling request formatting, authentication header injection, response parsing, and connection management. It uses axios for HTTP transport and implements a challenge-response authentication mechanism where API keys (generated via Anytype Desktop or CLI) are injected as Authorization headers on every request.

The command-line interface provides two primary functions: (1) authentication setup via `anytype-mcp auth` which guides users through generating API keys via Anytype Desktop and configuring environment variables, and (2) server startup via `anytype-mcp start` which initializes the MCP server and binds it to stdio for communication with AI assistants. The CLI abstracts away configuration complexity and provides interactive prompts for first-time setup.

Exposes Anytype's search API endpoints through MCP tools, enabling AI assistants to perform full-text search across all objects globally or within specific spaces. The search capability supports query parameters for filtering by object type, tags, and properties, returning ranked results with metadata that AI assistants can use to understand context and relationships within the knowledge base.

Enables AI assistants to create new objects in Anytype with specified types (e.g., Document, Task, Person) and templates, set properties and relationships, and organize objects into lists. The capability maps Anytype's object model (where each object has a type, properties, and relationships) to MCP tool parameters, allowing AI to construct complex knowledge structures through natural language instructions.

Exposes Anytype's space and collaboration management APIs through MCP, allowing AI assistants to create spaces, invite members, manage permissions, and configure space settings. Spaces are Anytype's unit of collaboration and encryption — each space is independently encrypted and can have different members with different permission levels.

Allows AI assistants to manage object properties (custom fields, metadata) and tags through MCP tools. Properties can be of various types (text, number, date, relation, etc.) and are defined at the object type level. Tags provide a lightweight categorization mechanism independent of the type system, enabling flexible organization without schema changes.

Provides AI assistants with access to Anytype's type system and template library through MCP tools. AI can discover available object types (Document, Task, Person, etc.), their properties and constraints, and available templates. When creating objects, AI can apply templates to inherit predefined properties and structure, ensuring consistency across the workspace.

Enables AI assistants to create and manage lists (collections of objects) and establish relationships (links) between objects. Lists are filtered views of objects matching certain criteria, while relations are explicit links between objects that can be traversed. AI can create lists based on object type or properties, add/remove objects from lists, and establish bidirectional relationships.